About this integration
Local-first memory — 100+ tools, 99.2% LongMemEval-S retrieval@10, hybrid search, GDPR, no cloud.
- Transport
- stdio
- Authentication
- none
- Initial setup
- none
- Runtime
- unattended
- Evidence
- documented
- Version
- 2026.9.14.4
- Package
- m3-memory
- Last compatibility test
- Not independently tested
Connect your agent
m3-memoryExisting Python package and repository identity were reconciled against the frozen listing. Publisher documentation reviewed only; no package, local model, database or endpoint was executed. The server persists potentially sensitive agent memory and offers deletion, synchronization and optional model/database integrations, so storage boundaries and tool permissions require operator review.
Capabilities: Store and retrieve agent memory, Run hybrid keyword and vector search, Track bitemporal history and supersession, Coordinate shared agent tasks, Run optional background enrichment
Connected profiles
Additional details
io.github.skynetcmd/m3-memory
Source ↗ · Checked 2026-09-172026.9.14.4
Source ↗ · Checked 2026-09-17CC0-1.0; package licenses are separate
Source ↗ · Checked 2026-09-17pypi
Source ↗ · Checked 2026-09-17m3-memory
Source ↗ · Checked 2026-09-192026.9.19.0
Source ↗ · Checked 2026-09-19The publisher explicitly documents an autonomous Cognitive Loop that runs in the background for deferred enrichment and a local MCP server process; this was not independently executed.
Source ↗ · Checked 2026-09-19The publisher states that the base local deployment needs no account or API key and makes no outbound calls; optional cloud endpoints and PostgreSQL deployment are separate configurations.
Source ↗ · Checked 2026-09-19The manual MCP configuration launches the installed m3 console script as the memory server; the same local store remains accessible through the CLI if a stdio session drops.
Source ↗ · Checked 2026-09-19Publisher documentation reviewed; package and integration endpoint not executed or independently security-audited.
Source ↗ · Checked 2026-09-19